Go 24/7
If you can leave your PC on 24 hours a day instead of 8-12 hours (or even less) then you'll generate more points and advance faster. That's pretty simple
The negatives to 24/7 folding are noise and the electric bill. Since you're at SPCR, the noise issue is pretty well covered here. As far as the electric bill goes, your best bet is to get as many points per kilowatt-hour as possible. My (completely unsubstantiated) view on that is that a PC running a 1.6GHz CPU doesn't draw much less power than one running a 3.2GHz CPU, so you might as well get the fastest box you can afford. I could be wrong, though, and you might get a better price/performance ratio from a slower low power CPU.
Start/Upgrade Early
Right now (7apr2004 - 17:00EDT) you need about 6200 points to get onto the first page (top 100) in the SPCR listings at folding.extremeoverclocking.com In a few hours it will take more points. It will never take less points. The sooner you can generate points, the better. If you have another PC kicking around that you've been meaning to put F@H on, do it now because putting it off loses points that you'll never get back. Life is like this in general, and folding is no exception.
Get more machines folding
This one seems pretty obvious, but what can you expect if you bring more machines on-line? The only folder whose setup I know well is me (though others have posted detailed descriptions of their setups) and I'm not 100% sure what's going on with my setup because half of it is at someone else's house. At the moment the EOC stats show me as green with 350 points/day on average. That is the combined effort of the following machines:
- Athlon XP 2000+ laptop on ~18-20 hours/day (off during commutes).
Athlon XP 3200+ (overclocked 2500+) on 24/7
Athlon XP 2400+ on 24/7
Pentium III 650 laptop on 24/7
Athlon XP 3200+ (OCed 2500+) on most of the time
Athlon XP 3200+ (OCed 2500+) on most of the time
Pentium IV 2.0GHz on unknown amounts of time
So that's what it takes for me to average 350 points/day. You might be able to get the same results with 3-4 Pentium IV hyperthreaded CPUs, but I'm not sure. At 350 points/day it would take about 3 weeks to get the 6200 points needed to be in the top 100 (provided I started 3 weeks ago.)
Right now the EOC stats show MikeC (mike_chin) as blue with 286 point/day on average. The Stanford folding stats page says that Mike has had 9 CPUs active within the past week. So with 2 more CPUs, Mike has generated 64 fewer points/day on average. There are a number of things that could account for that. Mike's machines could be running slower CPUs. Or they could be on less than 24/7. Or he could be running applications on them that use significant amounts of CPU power and leave less for folding. Or he could be getting work units that generate fewer points/day than then ones I've been getting. There are lots of variables.
So how about a single machine? I'm writing this post on the XP 2000+ laptop. Right now it's grinding away on a 70.9 point Tinker work unit (ugh!) Electron Microscope says that I'm getting a pathetic 36 points/day from this WU. At that rate, it would take about 6 months to generate the 6200 points needed to get into the top 100 (again, having started 6 months ago so that 6200 was still the number needed.) Fortunately, though, I don't get a lot of Tinkers (excepting the recent Tinker flood) on this machine and my point output is generally higher.
According to the EOC stats, my last full week of production netted me 2608 points from 42 WU. That's an average of 7 WU and 372 points every day. Breaking it down further, it means that I averaged one 53 point WU per day from each of my 7 machines, from the fast XP 3200+ boxes down to the slow P-III 650. So my 53 point/day average would have a single machine taking about 4 months to get to todays magic 6200 point number. Two machines with this production would get you there in 2 months, and four machines starting just one month ago would now have you in the top 100.
Borg
So we can see that more is better here, but more machines also means more electric bill. If "borging" (from the Star Trek TV series - The Borg were a organic/mechanical race who went around "assimilating" other people into their large collective being) is an option, then DO IT! If you have permission to put Folding@Home on other people machines and have them fold for you, you get more points and somebody else pays the electric bill. That's why I don't really know what's happening with 3 of the machines that are folding for me. They're not in my house. Do make sure to get permission, though. Your employer is not likely to look favorably to you installing software on their PCs that isn't for work purposes. People get fired over this. Getting permission in writing may be a good plan.
Optimize your settings
This has been debated a number of times here. The general consensus seems to be that you should do the following:
- 1. Set the client to do Folding@Home WUs instead of Genome@Home or whatever random chance gets you.
2. Use the -advmethods flag to have the client try to use the servers that have more of the higher-point-value Gromacs WUs.
3. Use the -forceasm flag to force the use of assembly language optimizations.
4. Use the -forceSSE flag if you have an Athlon CPU that has SSE.
Be PERSISTENT!
If you just keep on folding, you will move up. There are currently plenty of people in the top 100 who are generating zero points per day. Those are spots that are waiting for the folders who are persistent and keep generating points, no matter how many. The old "Tortoise And The Hare" story applies here. The Tortoise may not win the whole race, but he will certainly come in ahead of a Hare who takes off with a speedy start and then stops generating points. You can build a huge folding farm, Borg dozens of machines, optimize your setup, and fold 24/7, but if you let it all drop and you stop generating points, you'll eventually get beat by somebody with a single, slow system. You have to be persistent.
That's a good start. I certainly hope that others will add suggestions and advice.
Fold on,
Scott